PDA

View Full Version : UnrealMark UT2K4 performance thread for A64 (low end to hi end welcome)


Sazar
06-29-04, 07:44 PM
ragejg sez "splitting threads sux!!"

This once was a post by saz in the A64 Club thread, but I accidentally moved it over here when I split the thread to move me and Gator's stuff over here... SORRY SAZ!!!

So, seein's how this is the first post in the thread, lemme explain... Gator and I are looking at UT2k4 performance on a few different angles, CPU/vidcard/cache /RAM...

So, regardless of your vidcard, if you have an A64 of some sort, and UT2k4, then read on in this thread, and give some input please.

Gator
07-28-04, 02:18 PM
...
GATOR:

Install Umark (www.unrealmark.com)

Let me know what detail levels you want me to run at and stuff, and what botmatches to start etc. and I'll run off some benchies for ya (amd64/3000+/512mbDDR400/ti4400)...

:)

Try out DM-Antalus and DM-Deck17
12 bots
1024x768
no aa/af
highest details & textures
physics lowest

Let me know if these settings are good for you, or if you have a different suggestion and we'll compare notes :)

ragejg
07-28-04, 10:24 PM
@ your settings you specified:

http://www.nvnews.net/vbulletin/attachment.php?attachmentid=7208&stc=1

:)

ti4400 @ 300/600

cpu @ stock

memory: 512mb cas3, bt still pulls like 3100mb in sandra

Gator
07-30-04, 10:56 PM
UT2004 benchmark comparison:
DM-Antalus & DM-Deck17; 12 bots; 1024x768; no aa/af; highest details & textures; physics lowest

Ragejg - A64 3000, TI4400, 512mb PC3200 cas3 DDR
DM-Antalus = 62.817528 avg fps
DM-Deck17 = 86.436546 avg fps

Gator - A64 3200, FX5900XT, 1024mb PC3200 cas3 DDR
DM-Antalus = 93.756569 avg fps
DM-Deck17 = 97.856430 avg fps

Interestingly while my Deck17 score was not much over Ragejg, my Antalus score was about 50% faster which is a very impressive. I'm not entirely sure why there is such a difference, but unfortunetly Umark is also not an ideal benchmark since it does not move around the map in a fixed way like 3dmark. So if you bench yourself over and over on the same map, you would almost always get a different result. Another thing to bring into question is just how much my 1meg L2 cache improved my performance over your 512K cache, especially since our CPU's are the same clock speed. And while I really like my FX5900XT, I question if it was worth it when my primary interest was improving UT2k4's performance. Just the Athlon64 might have been enough to please my performance appetite. I'll be eager to see how it handles DoomIII, an Opengl game. But of course, the GF4TI will most definetly fall short for any DX9 games like HL2.

ragejg
07-31-04, 01:23 AM
Gator, within the week I'll have some 6800NU benchies to add, as well as some from a 5800U which is coming back to me soon. ;) I think the 5800U will post similar results to the 5900, and the 6800NU, well, I think it'll best the other cards by ~10fps... Add some AA and AF though, and, um, things might look a little different. :D

Gator
07-31-04, 02:36 PM
Gator, within the week I'll have some 6800NU benchies to add, as well as some from a 5800U which is coming back to me soon. ;) I think the 5800U will post similar results to the 5900, and the 6800NU, well, I think it'll best the other cards by ~10fps... Ad some AA and AF though, and, um, things might look a little different. :D

Looking forward to seeing the numbers. :) It'll be nice to see how 6800NU competes with the FX5900XT in a DX8 game. I rarely buy any games unless it holds my interest like the Unreal series does, so I expect Doom3 to be the last game I'll buy for a long time. That being the case, the 6800 series is of no use to me since the FX5900XT should be plenty adequate. But definetly post those numbers, and the FX5800U as well. And I thnk you're right, the FX5800U should be on par with the FX5900XT. The 5800U has higher clock speeds, but the FX5900XT has the 256bit-memory bus which makes all the difference ;)

ragejg
08-03-04, 05:14 PM
Ok, some 6800NU numbers...

same rig setup as ti4400 setup (3000+ 512k @ stock/512mb cas3 DDR @ 200)

@ stock card speed:
Antalus: 86.853523
Deck17: 94.872955

@ 378/805 (first autodetect with this card):
Antalus: 87.044800
Deck17: 95.058159

Now for a bump to 209fsb (HTT):
@ 400/850 (highest OC it can handle so far):
Antalus: 89.087395
Deck17: 97.487289

On to 220, where I ran into trouble, being able to run the benches with the card @ stock, but was unable to get the same OC out of it... so the following figure is @ stock card speed:
Antalus: 93.830856
Deck17: 102.649834

:)

Gator
08-04-04, 07:13 AM
http://www.nvnews.net/vbulletin/attachment.php?attachmentid=7325&stc=1

Some interesting results, it appears the 6800NU does not overpower the FX5900XT in UT2K4, a DX8 game. The FX5900XT even outperforms the 6800NU overclocked to 400/850! The only time the 6800NU marginally outperform the FX5900XT was when it's Athlon64 3000 was overclocked to a 3200 Newcastle level. This is disturbing news for people buying the 6800NU to replace an FX card if there primary interest is DX8 games and not DX9.

Another interesting question to ask is if my 1meg L2 cache and/or 1gig DDR memory helping the FX5900XT pull ahead?

Ragejg - A64 3000, TI4400, 512mb PC3200 cas3 DDR
DM-Antalus = 62.817528 avg fps
DM-Deck17 = 86.436546 avg fps

Gator - A64 3200, FX5900XT, 1024mb PC3200 cas3 DDR
DM-Antalus = 93.756569 avg fps
DM-Deck17 = 97.856430 avg fps

Ragejg - A64 3000, 6800NU, 512mb PC3200 cas3 DDR
DM-Antalus = 86.853523 avg fps
DM-Deck17 = 94.872955 avg fps

Ragejg - A64 3000, 6800NU @ 400/850, 512mb PC3200 cas3 DDR
DM-Antalus = 89.087395 avg fps
DM-Deck17 = 97.487289 avg fps

Ragejg - A64 3000 @ 2.2ghz, 6800NU, 512mb PC3200 cas3 DDR
DM-Antalus = 93.830856 avg fps
DM-Deck17 = 102.649834 avg fps

ragejg
08-04-04, 08:14 AM
Gator, I'll have a gig of memory soon enough, so we'll be even there... Also, the 6800NU doesn't make it's case @ 10x7, no AA/AF, it shines @ that rez with HEAVY AA/AF, and 12x10 with moderate AA/AF...

More on this later... :)

ragejg
08-04-04, 08:16 AM
Actually, yeah, Gator, you should run those same benchies @ 10x7 with 4XAA/8XAF...

After that, 12x10, no AA/no AF, then 12x10 4xAA/8XAF...

Gator
08-04-04, 08:29 AM
Actually, yeah, Gator, you should run those same benchies @ 10x7 with 4XAA/8XAF...

After that, 12x10, no AA/no AF, then 12x10 4xAA/8XAF...

You're right, AA/AF is where the 6800NU will likely shine. I'm sure my AA/AF scores will stink compared to yours. Now I can't bench 12x10 because my LCD only supports up to 10x7. So tonight lets compare 1024x768 4XAA/8XAF, and if you think of any other AA/AF settings you want me try then post here. And be sure to include benches for all hardware at stock speed, as well as overclocked if you want. And as before: DM-Antalus & DM-Deck17; 12 bots; 1024x768; no aa/af; highest details & textures; physics lowest

ragejg
08-04-04, 09:38 AM
And as before: DM-Antalus & DM-Deck17; 12 bots; 1024x768; no aa/af; highest details & textures; physics lowest

You mean 4xAA/8xAF? :p

Gator
08-04-04, 11:32 AM
You mean 4xAA/8xAF? :p

opps, yes that's what I meant...post corrected... ok so to confirm:
DM-Antalus & DM-Deck17; 12 bots; 1024x768; highest details & textures; physics lowest; 4xAA/8xAF

Gator
08-05-04, 05:35 AM
DM-Antalus & DM-Deck17; 12 bots; 1024x768; highest details & textures; physics lowest; 4xAA/8xAF
A64 3200, FX5900XT, 1024mb PC3200 cas3 DDR
DM-Antalus = 67.586484 avg fps
DM-Deck17 = 69.767044 avg fps

boy AA/AF sure does give my fps a beating :bash:

ragejg
08-05-04, 07:19 AM
DM-Antalus & DM-Deck17; 12 bots; 1024x768; highest details & textures; physics lowest; 4xAA/8xAF
A64 3000 @ 209x10, 6800NU @ 325/700, 512mb PC3200 cas3 DDR
DM-Antalus = 81.091209 avg fps
DM-Deck17 = 88.468887 avg fps

DM-Antalus & DM-Deck17; 12 bots; 1024x768; highest details & textures; physics lowest; 4xAA/8xAF
A64 3000 @ 209x10, 6800NU @ 410/860, 512mb PC3200 cas3 DDR
DM-Antalus = 88.511925 avg fps
DM-Deck17 = 89.091667 avg fps

Sazar
08-05-04, 09:45 AM
I started this thread :confused:

-edit-

ah nm... read rage's psuedo-edit :D

Gator
08-05-04, 01:08 PM
nice! Clearly the 6800NU is much better at AA/AF than the FX5900XT. It doesn't appear that overclocking the video card does much though, so it might not even be worth overclocking it in the future.

ragejg
08-05-04, 01:10 PM
... I've done some testing @ 12x10 with high AA and AF, and the card is helped out quite a bit more at that rez... I'd iagine that 16x12 would experience even more improvement. :)

Gator
08-05-04, 01:11 PM
... I've done some testing @ 12x10 with high AA and AF, and the card is helped out quite a bit more at that rez... I'd iagine that 16x12 would experience even more improvement. :)

Nice! I wish I could compare numbers at that res but I can't. But post yours it'd be good to see.